Failed with 'CUDA_ERROR_OUT_OF_MEMORY'
miraclezqc opened this issue · 0 comments
miraclezqc commented
Hello!I tried to compile the "open earth compiler" project, and set MLIR_CUDA_RUNNER_ENABLED in llvm-project/mlir directory to make sure that have the libcuda-runtime-wrappers, then I followed the commands of pipeline and And encountered the following error when running the main.cpp
'cuMemAlloc(&ptr, sizeBytes)' failed with 'CUDA_ERROR_OUT_OF_MEMORY'
'cuMemAlloc(&ptr, sizeBytes)' failed with 'CUDA_ERROR_OUT_OF_MEMORY'
'cuStreamCreate(&stream, CU_STREAM_NON_BLOCKING)' failed with 'CUDA_ERROR_ILLEGAL_ADDRESS'
'cuModuleLoadData(&module, data)' failed with 'CUDA_ERROR_ILLEGAL_ADDRESS'
'cuModuleGetFunction(&function, module, name)' failed with 'CUDA_ERROR_INVALID_HANDLE'
'cuLaunchKernel(function, gridX, gridY, gridZ, blockX, blockY, blockZ, smem, stream, params, extra)' failed with 'CUDA_ERROR_INVALID_HANDLE'
'cuModuleUnload(module)' failed with 'CUDA_ERROR_INVALID_HANDLE'
'cuMemFree(reinterpret_cast<CUdeviceptr>(ptr))' failed with 'CUDA_ERROR_ILLEGAL_ADDRESS'
'cuMemFree(reinterpret_cast<CUdeviceptr>(ptr))' failed with 'CUDA_ERROR_ILLEGAL_ADDRESS'
My environment and command:
llvm-project commit id: e59d336e75f(build with gcc & g++)
g++ -O3 -c main.cpp -o main.o -fPIE -I/share/software/cuda-11.1/include/
g++ main.o laplace.o -o main -fPIE -L/share/userfile/user/mlir/lib -lcuda-runtime-wrappers -lcuda -L/share/software/cuda-11.1/lib64/ -lcudart
I would to ask how to solve this problem?